Pricing model for the Italian electricity market

Abstract

This master thesis has as a main objective the prediction of expected zonal prices in the Italian Electricity Market based on a pricing model according to the thermal gap. This objective will enable to obtain appreciated information of the market and to understand the functioning of it. The model can be defined as a deterministic, fundamental and hybrid model. Training data consists of data from 2016 to 2017 while forecasted data covers the period from 2018 to 2019. Within the first months of the forecasted data, the period from January to May will be considered as test data used in the Validation process. The general methodology is a model based on a stack model, this means the demand shall be estimated and then it is covered with production, ordered according to its marginal cost. In this model, demand is covered until the thermal gap (or thermal production) is left. As supply curves for thermal plants are unknown, an algorithm is developed in order to build these curves based on historical data, therefore these curves will relate the thermal gap with the price. The algorithm abovementioned, replicates the price formation in each bidding zone and has as an output a coefficient for each virtual plant to be added in its cost formula. This coefficient allows to capture the strategy of the different agents and will be different for each day and period classification. Thus, the model is not a formal stack model neither a static one since the strategy of the agents is captured with these coefficients and the cost formula for thermal technologies varies with commodities price in the market. The validation process provided positive results and showed that the model adjusted its outputs according to actual data. In addition, in this analysis, residuals were close to zero which is a condition for a good accuracy in the forecast. If results are compared with the expectations of forward markets, it was showed how similar the values were. Therefore, one can conclude that results and the model development were satisfactory, considering that this model was developed from scratch where great scope for improvement exists. Some of the improvements are included in the Section Future Works.